Application Research Of Swarm Intelligence Algorithm On Block Matching Motion Estimation

As an important research content of video processing, motion estimation is widelyused in video compression, video surveillance, target tracking and computer visionetc. Method for motion estimation consists of flow equation method, pixelrecursive method, block matching method and grid method. Among them, whichbased on block matching has gradually become the mainstream research methodfor motion estimation due to its simple ideology and easy implementation. But inthe previous research of block matching motion estimation, the fixed template isusually adopted to search the best matching block. This kind of method is based onthe matching error decreases monotonically hypothesis which is often inconsistentwith the actual problem, and will cause high matching errors. In recent years, withthe development of research on the swarm intelligence algorithm, a new researchidea is proposed as a method of motion estimation. In this paper, several swarmintelligence algorithms with superior performance are chosen to beimproved according to the features of video sequences, and then applied to blockmatching motion estimation. The proposed methods were proved that they haveovercome the defect of premature convergence and easy to fall into localoptimum, and accordingly reduced the matching error, improved the accuracy ofmotion estimation finally.In this paper, the main research results are as follows:(1) Aimed at the selection of the two random numbers of velocity’s updatingformula, an advanced particle swarm optimization algorithm based on good pointset is proposed. The new method uses a good point set to improve the constructionof the two numbers, and then applied to block matching motion estimation. Theresults of experiment show that the advanced algorithm has accelerated theconvergence speed, thus the search precision of its application in block matchingmotion estimation has improved.(2) For the defects of premature convergence and easy to fall into localoptimum of artificial bee colony algorithm, an advanced algorithm with theintroduction of good point set and turn process of monkey algorithm is proposed.The new method constructs the initial population with good point set, and introduces the turn process after the updating of the observing bee, and thenapplied to block matching motion estimation. The results of experiment show thatthe proposed algorithm has good ability of fast convergence, it can avoid trippinginto local optimum, and it also has strong robustness, thus the search accuracy ofits application in block matching motion estimation is signification higher than thatof other methods.(3) Aimed at the slow convergence speed and the insufficient local ability ofglowworm swarm optimization algorithm, an advanced algorithm based onsingle-dimension disturbance strategy is proposed. The new method introduces asingle-dimension disturbance process after the updating of position, and thenapplied to block matching motion estimation. The results of experiment show thatthe comprehensive performance of the proposed algorithm is better than previousmethods for most of the video sequences, especially for the video sequences withviolent motion, and the search accuracy has improved obviously. The computationcomplexity of the proposed algorithm is slightly higher than that of classicalmethods but still far lower than that of full search method.